Flow Modeling Provides Data-Filled Window into Gas Turbine, Plant Performance

At the 2025 Western Turbine Users Inc. (WTUI) conference in Long Beach, CA, Engineering Manager Matt Gentry of Airflow Sciences Corp. presented on flow modeling for aero power plant performance optimization and addressed how it can identify, solve, and optimize parameters such as flow, temperature, ammonia, NOx, and pressure drop.

“Flow modeling is an important component of the engineering toolbox,” Gentry said. “Both computational fluid dynamics (CFD) and physical-scale modeling can provide detailed predictions of flow, pressure, ammonia, and temperature distributions throughout a gas turbine power plant. These predictions allow for performance optimization without the need for a time-consuming and expensive trial-and-error approach to solving problems in the field.”

Gentry said many flow-related parameters are inter-related:

• Ammonia non-uniformity can cause pressure loss and heat-rate issues
• Poor velocity distribution can result in skewed heat transfer at tube banks
• Tube bank, catalyst, and ammonia injection grid (AIG) seals/baffles may seem simple, but if they are not maintained, they can have a significant negative impact

TURBO: How does flow modeling optimize a plant’s performance?

Gentry: Flow modeling is performed on all modern plants in the design stage prior to construction. With millions of dollars on the line, asset owners want some verification that the plant will operate as expected; similarly, after years of operation, it makes sense to ensure the plant is operating optimally. Flow modeling is an important tool for any process that utilizes fluid flow or heat transfer.

Flow modeling can provide value for a gas turbine power plant by diagnosing performance issues, such as excessive pressure drop, poor heat exchange, suboptimal emissions, etc. After identifying these issues, modeling is essential to quickly enacting and verifying any potential design changes; it increases plant efficiency and uptime. When every dollar counts, the savings add up.

TURBO: What does flow modeling entail?

Gentry: Flow modeling typically refers to two different tools: computer flow modeling, referred to as CFD, uses high-speed computing resources to predict a plant’s gas-flow parameters. A full-scale computer-aided design model of the plant is constructed, and the model is discretized into many small computational volumes or cells. Using a CFD model, precise gas turbine exhaust-flow data can be implemented, allowing for highly accurate predictions of the flue gas conditions downstream.

In physical-scale modeling, a small-scale version (1:10, 1:12, or 1:16 are common) of the plant is constructed. The model is connected to a fan in a laboratory setting and data can be measured. The model is run at ambient conditions and calculations are utilized to ensure that the results are in the correct Reynolds number regime and are thus applicable to the full-scale hot-gas flow. Some simplifications are required, especially in regard to the exhaust flow distribution, but some clients prefer the physical model as it is tangible and visual. With 3D animations on high-speed computers, a CFD model can give the client much deeper insights into the complex flow field.

TURBO: Does flow modeling for aeroderivative assets differ from single- or combined-cycle assets? How so?

Gentry: Nope! There is no difference between the techniques required for aeroderivative, simple cycle, or combined cycle. The varying plant types may have different goals or challenges, but the same flow modeling techniques can be applied to any gas turbine asset. Fluid flow or heat transfer? Modeling can help with plant issues.

TURBO: How can CFD redesign and improve ammonia injection grids?

Gentry: CFD is already being used to design and improve AIGs all the time. Older AIG systems that were not modeled prior to commissioning could benefit from modern analysis. CFD can be used to optimize AIG performance in the following two respects:

• A detailed model of the AIG itself accounts for the internal flow and allows for an accurate prediction of the nozzle-to-nozzle flow distribution. If required, design modifications can be sought to achieve the uniform distribution that is often critical for downstream mixing.
• A model of the gas turbine exhaust ductwork will predict the distribution of ammonia downstream after it has exited the AIG. All catalyst vendors require certain ammonia distribution uniformity for their performance guarantees to be valid. The full-duct CFD allows the ammonia distribution at the catalyst inlet to be calculated. If needed, mixing devices can be designed and validated using the duct CFD to improve and optimize the ammonia distribution, improving catalyst performance and reducing emissions.

In a case study, two combined-cycle plants were handling exhaust from 501F gas turbines. The plant was having issues with ammonia distribution and NOx catalyst performance. In addition, the heat transfer tubes required frequent cleaning, as ammonia salts were accumulating and significantly increasing the back pressure on the turbine.

CFD modeling was used as the analysis tool for this project. The baseline CFD results demonstrated poor ammonia distribution at the selective catalytic reduction (SCR) catalyst face. This matched the client’s anecdotal experience and indicated poor catalyst performance and ammonia slip. The CFD results indicated that non-uniform flow near the AIG was causing ammonia to be concentrated in recirculation zones. This was primarily due to an expansion of the heat recovery steam generator (HRSG) ductwork's cross-sectional area in the AIG region.

In flow modeling, a data metric known as root mean square (RMS) is often used: a determination of uniformity analogous to the standard deviation expressed on a percentage basis (standard deviation divided by the average value) but not limited to 100%. Very non-uniform distributions can have RMS values in excess of 100%. An RMS of 0% would indicate perfect mixing. The industry standard for acceptable “uniform” distribution is 10 - 15%. Catalyst vendors will specify an ammonia RMS requirement, which must be met to validate their removal guarantees. A gas turbine SCR often requires an ammonia RMS of 10% or less.

For this case study, the ammonia distribution at the catalyst face was 69% in the baseline case, significantly worse than the required 10%. As a result, a detailed design analysis was completed. In the final design, the non-uniform flow concerns near the AIG were corrected with the addition of flow control baffles to redirect flow inward, eliminating recirculation zones and reducing AIG bypass. In addition, the AIG was redesigned to include a local static mixer to improve ammonia distribution at the SCR. The mixer was predicted to have a very low-pressure loss (~0.1 inch H₂O), allowing for improved mixing with minimal impact on the overall back pressure. The position of the AIG was also moved further upstream, allowing for more residence time for mixing prior to the SCR catalyst.

The plant implemented the recommended design changes in 2018. They saw a small increase in pressure loss due to the new baffles (0.25-inch H₂O), but this pressure drop held steady over the next six months, indicating that the ammonia salt buildup on the tubes was not occurring. By improving the ammonia distribution, the plant saw improved emissions and lower ammonia slip. Their ammonia consumption was reduced by 20%, resulting in a significant operating cost reduction. As a result of the design changes, the ammonia RMS at the catalyst face was improved from 69% to 7% RMS.

Airflow Sciences Corp. followed up with this plant in January 2024 (6.5 years after the modifications). The plant engineer said: “We did see an immediate decrease in aqueous ammonia usage (~20%), improved NOx control, and a reduced rate of ammonia salt deposition on the HRSG tube surfaces over time. The bulk of salt deposition also appears to have shifted further back in the HRSG from module 5 to module 6. As you know, the reduced salt deposition directly affects heat rate, but there hasn’t been any reason for us to quantify the correlation. Since the project was completed, we have only cleaned the HRSGs once, and minimal debris was recovered. In our opinion, the project was a success.”

The use of flow modeling in this case allowed for a bespoke design, developed in less than two months. Utilizing the CFD model and high-speed computers, many different design modifications were analyzed in a short amount of time. In a collaborative process with the plant, the final design was selected based on performance improvements, as well the ease of installation.

TURBO: What issues do gas-fired units present for flow-modeling operations?

Gentry: Think of a 1,200° F tornado blasting out of a duct—that’s what’s being discharged from a gas turbine. Properly characterizing this swirling flow presents a challenge for flow modeling. This is particularly true for physical-scale modeling, where matching the exact profile is impossible and simplifications must be applied. This is less of a factor for CFD models, where the velocity data from the gas turbine manufacturer specifications can be applied as a boundary condition; however, the outlet profile can vary, sometimes substantially from load to load. Thus, if a CFD modeling project is only being run at full-load operating conditions, the flow control and mixing device design must be sufficiently robust to withstand some variation in the velocity profile.

TURBO: What are some performance degradation issues that can be identified by flow modeling?

Gentry: Any flow or heat problem can be investigated. Just to name a few: excessive pressure loss, poor tempering air mixing, poor ammonia mixing or catalyst performance, poor heat transfer, excessive ammonia slip, catalyst degradation due to excessive temperatures, and excessive noise or vibration.

TURBO: What are the latest technological advancements in flow modeling technologies?

Gentry: Ever-increasing computer performance is a boon to CFD modeling. Numerical models require significant amounts of random access memory (RAM) and fast processor speeds. As computers improve, the CFD modeling projects can become more detailed (denser mesh of cells), allowing for more accurate results. Faster processing speeds also translate to quicker project turnaround, as modeling simulations can be completed in shorter durations. The basics of physical modeling have not changed too significantly since the 1970s, but new manufacturing tools allow for faster or more accurate model construction. Today's 3D printers can fabricate a complicated mixing device over a weekend, compared to days of construction in a fabrication shop.